Comment by regal | Hacker News Reader

Comment by regal | Hacker News Reader

Parent Full thread

Had the opposite experience - no change in loading speeds when our ~600K visitor / month site switched to CloudFlare in March of this year, but we soon started experiencing long website downtimes where our server was running fine but CloudFlare was not serving our site, despite paying $200 a month for "always online", which the site clearly was not. All IPs were whitelisted with our host and everything else CloudFlare recommended; all CloudFlare would do was examine the site after it had come back online (following hours of downtime) and say, "Everything looks fine to us!"

We finally left, after months of this, and have had no problems with downtime since. I really wanted CloudFlare to work - I was really excited about it when I signed us up. But at least for a bigger site with heavier traffic that relies on being up as much as possible, I can't say I'd recommend it until it straightens out its downtime issues (especially when paying for "always online").

I keep hearing sporadic reports like this. The big thing with CloudFlare is you are putting them as the first hop in reaching your site. So they are an additional point of failure. Of course, it's not a zero-sum game, they could also end up increasing your uptime overall, and in many cases I believe that's the case.

Particularly for relatively low volume sites which have a short burst in traffic on occasion, CloudFlare can keep those sites running during the peaks.

I think the most important thing is transparency and correct expectations. If they set clear expectations, and they are transparent about how well they are meeting them, then it just comes down to delivery.

I found their status dashboard here: https://www.cloudflare.com/system-status. Unfortunately it doesn't show much long-term historical performance, it would be nice to see 30 days even 180 days of performance history to really evaluate them.

regal, did you find that when you had downtime on your site that it was reported in their status dashboard, and that was an accurate depiction of the service they provided? I think the worst-case scenario is getting hit with unreported downtime, because that brings up all sorts of questions.

I think the most important thing is transparency and correct expectations. If they set clear expectations, and they are transparent about how well they are meeting them, then it just comes down to delivery.

Agreed. So long as a customer knows what he/she's signing up for, and gets that, everything's fine. I might have misread what the "99.99% uptime guarantee" was supposed to be for and gotten too excited about it / taken it too seriously when I first signed up, or maybe this is for something else that's too complicated for a part-time tech guy like me to understand.

When I'd log in when the site was down, half the time CloudFlare would have the green arrow next to the site with a "Site Online" type indicator; other times it'd have the brown dot-dot-dot "Site Offline" indicator. I'd confirm numbers on this but apparently the service doesn't save this or makes it no longer available to you on account termination. Pingdom Tools would report the site as down, and when visiting the site, it wouldn't load, or would take 10+ seconds to load. There would also frequently be a "This website is offline; no cached version is available" page from CloudFlare when trying to load the site, even on the homepage, despite the guarantee to supposedly be saving and serving cached copies of the site in the event of downtime (and despite that being what I thought we were paying for, mainly) - sometimes those cached copies would show up too; though more often, there'd just be this page:

http://image2.romantika.name/2013/01/cloudflare-website-offl...

I opted to use CloudFlare almost as soon as they launched and they indeed had some uptime issues in the beginning. But I stuck with them and I haven't had any issues for a long while now even at 180M page views / month.

There was one time where I had just upgraded a server at a host and load averages just spiked. Host techs were clueless as well as the server admin I hired to diagnose the issue. They surmised that it was a DDoS attack and to turn off CloudFlare because it might be a cause (what?). Well I'm not a server admin so what do I know and I asked CloudFlare about it. They said there were no problems on their end. After a month of stressful, intermittent downtimes, I decided to just switch to Softlayer, and lo and behold, the issues went away. Turned out that one of the SSDs in the RAID array was dying but the techs at the other host just never bothered to look at it.

If you ever do figure out the issues, try giving them another shot. Their features are excellent and I hate to say it but my applications are now so dependent on them to the point where important parts will break if I even try to move to a competitor (and there are none).

If you read this... please reach out to me if willing via the email in my profile here. Would like to dig into what happened.

Reply on news.ycombinator.com